摘要
[目的/意义]对微博消息进行热点话题挖掘,进而从海量微博文本中实时找出用户关注、讨论的热点事件,是进行舆情监测、应急管理的基础。然而,现有微博热点话题检测研究却大多忽略了不同影响力用户对话题产生及传播的作用,并且检测结果直观性较差。针对此问题,提出了基于用户影响力的热点话题检测方法。[方法/过程]首先识别用户特征要素,构建用户影响力模型,计算用户影响力;然后,综合考虑主题词影响力、影响力增长速度和增长斜率,提出基于用户影响力的微博热点话题主题词抽取方法,抽取主题词簇;之后,识别核心主题词并进行热点话题关键词抽取。最后,通过实验验证方法的有效性。[结果/结论]实验结果表明:基于用户影响力的热点话题检测方法能够有效识别并直观表达出检测时间窗口内的典型热点话题;该方法能有效提升实证性热点话题识别效率,减少娱乐性热点话题的识别;通过对不同时间窗口内同一话题的关键词抽取,可以实现对相应话题的热点跟踪。
i Purpose/Significance] Hot topic detection could provide basic support for public opinion monitoring and emergency manage- ment. However, most existing researches on hot topic detection usually ignore the impact of micro-blog users in the process of mieroblog generation and spread. Besides, the results of those researches lack of visibility. To solve these problems, a user influence-based hot topic detection method in micro-blog is put forward. [ Method/Process] Firstly, a user-influence model is proposed to compute the influence of different users. Then, taking feature words＂ influence, influence＇s growth rate and influenee＇s rising slope into consideration, the feature words extraction is performed. After that, the core feature word is identified and the keywords of the hot topics are extracted based on the core feature word. Finally, the effectiveness of this method is verified by experiment. [ Result/Conclusion] The results showed that： the hot topic can be recognized by user influence-based hot topic detection method effectively, the method can recognize empirical topics more effectively, the method can achieve the corresponding topic＇s tracking.
出处
《情报杂志》
CSSCI
北大核心
2017年第4期156-161,共6页
Journal of Intelligence
基金
辽宁省社会科学规划基金重点项目"突发事件网络舆情的动态监测与预警策略研究"(编号:L15AGL017)
国家自然科学基金项目"在线知识社区中社会系统与知识系统协同序化机制和规律研究"(编号:71573030)的研究成果之一
